Employment outcomes and policies in Sweden during recent decades *

Transcription

1 Employment outcomes and policies in Sweden during recent decades * by Anders Forslund * The report is prepared for the Expert Group of the Employment Commission. IFAU and Uppsala University. The author thanks members of the Expert Group of the Employment Commission for valuable comments. Any remaining mistakes are solely due to the author. IFAU Employment outcomes and policies in Sweden during recent decades 1

2 Table of contents 1 Introduction Swedish labour market outcomes during recent decades Skills upgrading institutions and policies Upper secondary vocational programmes Adult education (Komvux and Sfi) Adult vocational training programmes (Yrkesvux) Vocational introductory jobs (Yrkesintroduktionsanställningar) Vocational labour market training programmes (Arbetsmarknadsutbildning) Vocational college (post-secondary) programmes (Kvalificerad Yrkesutbildning and Yrkeshögskolan) Participation in adult education and training in Sweden Employment effects of the different programmes Measures of skills among adults and adult education Rehabilitation Policies for immigrant integration Skills upgrading institutions and policies a brief summing up Job-search incentives Income taxes Unemployment insurance Benefits for programme participants and newly arrived immigrants Social assistance benefits Sickness insurance Pensions Early retirement Child care Parental leave Job-search incentives a brief summing up Demand-side policies The public sector as an employer Policies to reduce labour costs Policies to reduce employer uncertainty Employment protection legislation and the use of temporary contracts Demand-side policies a brief summary Measures to improve matching Job-search assistance and monitoring search efficiency, intensity and search costs Concluding remarks IFAU - Employment outcomes and policies in Sweden during recent decades

3 1 Introduction Employment is the result of complicated processes which involve decisions to invest in different skills, decisions to apply for jobs, decisions to invest in production technologies and decisions to hire persons with given sets of skills. Each of these components of the processes are time- and resource-consuming activities, and all of them are influenced by different policy measures and the design of several institutions (private or public). In this report I provide a discussion of how employment in Sweden during recent decades has been influenced by different kinds of public policy. As a foundation for the policy discussion, I first give a brief description of the key features of recent developments in employment-related outcomes in the Swedish labour market. The discussion of policy on the labour-supply side focus on skill-upgrading policies and institutions, as well as on policies working through job-search incentives. I also cover policies working through labour demand and (although very briefly) policy measures affecting matching in the labour market. 1 2 Swedish labour market outcomes during recent decades The Swedish employment rate in 2017 was 81.8 per cent according to Eurostat statistics, almost ten percentage points above the EU average (72.2 per cent). Of the countries in the Eurostat data base, only Iceland (87.6 per cent) and Switzerland (82.1 per cent) had higher employment rates that year. The difference between Sweden and the EU average in employment rates can be decomposed into a difference in labour force participation rates and a factor reflecting differences in unemployment rates. Performing such a comparison immediately reveals that the higher Swedish employment rate almost exclusively reflects higher labour force participation. The difference in participation rates was 9.1 percentage points in 2017, whereas the difference in unemployment rates was 0.9 percentage points. 2 So, to get an understanding of the high Swedish employment rate, the descriptive evidence seems to suggest that we primarily should be interested in factors related to labour supply. 1 To some extent the distinctions between supply, demand, and matching are fuzzy. It is not, for example, evident if labour market training programmes should be counted primarily as supply or matching measures. 2 The employment rate equals the participation rate times (1 - the unemployment rate). IFAU Employment outcomes and policies in Sweden during recent decades 3

4 In Figure 1, the Swedish employment and participation rates for year-olds from 1986 to 2017 are plotted. Although the employment and participation rates were higher in 2017 than they were for a long period before this year, they were even higher before the deep Swedish recession of the 1990s. But in the present context, the most interesting feature of Figure 1 is that the development since around 2005 has been favourable with increasing employment and participation rates, with only a brief setback during the international financial crisis. There is an interesting difference between the 1990s crisis and the international financial crises in the adjustment patterns. The recession in the 1990s involved significant and long-run decreases in both employment and participation. In contrast, the Swedish response to the international financial crisis was short lived, and the decrease in employment was accompanied by only a minor decrease in participation year Employment rate Participation rate Figure 1: Employment rate and participation rate Note: Figures for population aged Source: Statistics Sweden, Labour force surveys. 4 IFAU - Employment outcomes and policies in Sweden during recent decades

5 The development of the unemployment rate (Figure 2) has not been equally favourable after the crisis of the 1990s, but unemployment is almost back to where it was before the international financial crisis year Figure 2: The Swedish unemployment rate Note: Figures for population aged Source: Statistics Sweden, Labour force surveys. The recent steady rise in employment and participation rates hides differences across different strata in the Swedish population. In Figure 3 we see the remarkable growth in employment and participation rates among the elderly (55+ years of age) during recent decades. Although employment and participation among the prime-aged has also increased recently, the increase is much smaller than among the old, as is evident from Figure 4, where employment and participation rates among persons between 25 and 54 years of age are displayed. 3 3 It is also striking how high the Swedish participation and employment rates among the elderly are in comparison with the EU average: in the first quarter of 2018, the Swedish participation and employment rates were 80.9 % and 77.2 % compared to 61.3 % and 57.8 % among persons aged according to Eurostat figures. IFAU Employment outcomes and policies in Sweden during recent decades 5

6 year Employment rate Participation rate year Employment rate Participation rate Figure 3: Employment and participation rates in 55+ ages Source: Statistics Sweden, Labour force surveys year Employment rate Participation rate Figure 4: Employment and participation rates in ages Note: Figures for population aged IFAU - Employment outcomes and policies in Sweden during recent decades

7 Source: Statistics Sweden, Labour force surveys. Outcomes for young persons are somewhat more complicated to describe in an informative way. High employment and labour force participation rates are not necessarily desirable we expect most young persons to be in education or training. Unemployment rates are commonly used to describe how young persons do in the labour market. In terms of youth unemployment rates, Sweden is not doing particularly well the Swedish youth unemployment rate in 2017 (17.8 per cent) was actually above the EU average (16.8 per cent) according to Eurostat numbers. However, youth unemployment rates are unfortunately not very informative when comparing the labour market prospects for young persons in different countries. Some unemployment differences do capture real differences in labour market prospects, but a potentially substantial share of the differences may instead capture institutional differences giving rise to pure book-keeping differences. Such institutional differences include the extent to which vocational training programmes at the high-school level are apprenticeship programmes (in which young persons are counted as employed if they are paid any wages) or not and whether or not university students receive study grants during vacations (in which case they rarely apply for jobs) or not. For this and other reasons it has become more common to compare inactivity rates (shares of young persons who neither study or work, NEETs ) among young persons. Comparing the Swedish NEET rate to the EU average gives a totally different picture than the comparison of unemployment rates: the Swedish NEET rate in 2017 was 6.1 per cent compared to the EU average at 10.9 per cent. In Figure 5 we see the NEET and unemployment rates among year-olds. The level difference primarily reflects that the denominators are different the denominator for the unemployment rate is the labour force, whereas the denominator for the NEET rate is the population. Unfortunately, the series are only available for a short period, but both measures suggest a trend wise improvement in the labour market prospects for young persons since the international financial crisis. 4 There are, however, groups of young persons for which labour market prospects are problematic, and for certain groups even increasingly so over time. An analysis of the 4 Remember, however, that it is not entirely clear that the youth unemployment rate in any meaningful way measures the development of the labour markets prospects for young persons. IFAU Employment outcomes and policies in Sweden during recent decades 7

8 labour-market entry of young persons in Engdahl & Forslund (2016) identifies three, partly overlapping groups of young persons with bad labour-market prospects: highschool dropouts, young persons with immigrant background and, especially, young disabled persons. It seems particularly problematic that the group of high-school dropouts is increasing and the group of young disabled both is increasing and doing worse over time. Another possible problem is that young swedes with tertiary education enter the labour market at above-average ages. This has for example been pointed out by the OECD (2008) and can be seen in the average age at which tertiary education graduates leave school in the latest available OECD comparison (for 2015), Sweden had the highest average graduation age, reflecting both a high average entry age and a long duration of the studies year Unemployment NEET Figure 5: Unemployed and NEETs years old Source: Statistics Sweden, Labour force surveys. In Figure 6 we look instead at the development of employment and labour force participation by gender. The levels of employment and labour force participation are slightly higher among males, but the differences become smaller over time, reflecting the fact that both participation and employment rates have grown faster among females. 8 IFAU - Employment outcomes and policies in Sweden during recent decades

9 Compared to the EU average, the Swedish participation and employment rates are higher both among males and females, but the difference is much bigger among females, which is evident from the figures in Table year Female employment rate Female participation rate year Male employment rate Male participation rate Figure 6: Employment and participation rates by gender, Note: Figures for population aged Source: Statistics Sweden, Labour force surveys. Table 1: Employment and participation rates in ages first quarter of 2018 in Sweden and EU 28 Employment rate (%) Participation rate (%) Females Males Females Males Sweden EU Source: Eurostat. In Figure 7, the development of employment and participation rates among natives and immigrants in Sweden aged are displayed. The gap in employment rates is just IFAU Employment outcomes and policies in Sweden during recent decades 9

10 below 20 percentage points and somewhat shrinking after the international financial crisis, but the employment rates have developed in rather similar ways (slowly increasing) for both strata from 2005 and on. However, the fact that the immigrant share of the population has grown over time means that immigration has contributed negatively to the development of the aggregate employment rate in a purely book-keeping sense. Notice also that the participation rate among immigrants, in contrast to the development among natives, is increasing rapidly over the period year Employment rate immigrants Participation rate immigrants year Employment rate natives Participation rate natives Figure 7: Employment and participation rates by place of origin , ages Source: Statistics Sweden, Labour force surveys. In Figure 8 we see employment and participation rates by level of educational attainment. Hardly surprising, employment and participation rates are monotonically increasing in the level of educational attainment. Perhaps a little more surprising, we see that for all but individuals with post-secondary education, employment rates decrease over time. Hence, the aggregate increase in the employment rate in a mechanical sense is due to 5 I show the development of participation and employment in prime ages to take some account of the different age distributions of natives and immigrants. 10 IFAU - Employment outcomes and policies in Sweden during recent decades

11 compositional change, where the level of educational attainment in the population in working age steadily has gone up year Employment rate primary Participation rate primary year Employment rate secondary Participation rate secondary year Employment rate tertiary Participation rate tertiary Figure 8: Employment and participation rates by level of educational attainment Note: Figures for population aged Source: Statistics Sweden, Labour force surveys. 3 Skills upgrading institutions and policies In the Swedish labour market model, adjustments to changes in demand have traditionally been accomplished primarily through employment changes (layoffs in cases with decreasing demand) and not through wage changes. In such a system, it is important that there exist institutions and policies to facilitate the skills upgrading necessary for occupational mobility. It is not a coincidence that vocational labour market training for the unemployed was an important part of the policy programme, among other things aiming at rapid structural change to promote productivity growth, lunched by the central 6 It is not too far-fetched to believe that the decrease in employment rates for those with lower educational attainment at least partly is a selection phenomenon the selection of individuals into lower educational attainment has probably meant selection on less marketable qualifications as the group has decreased over time. This is consistent with a growing unemployment rate among those with less than secondary education. IFAU Employment outcomes and policies in Sweden during recent decades 11

12 union of blue-collar workers in the late 1940s and early 1950s and that such training measures rapidly became the jewel in the crown in Swedish active labour market policies. Recent developments under the headings technical change and globalisation have probably further strengthened the need for skills upgrading for adults. Another factor also working to increase the need for adult training is the poor results in Swedish primary and secondary school, especially the fact that a significant proportion (around 20 per cent) of each recent youth cohort ends up without eligibility for any upper-secondary national programmes and, with a high probability, becomes high school dropouts. Finally, a significant share of the individuals in the recent large influx of immigrants have a poor school background, and hence, many of them are candidates for skills upgrading. Hence, we can identify needs for adult education and training for a number of reasons. So, is there supply to meet the needs for adult education and training? I start by presenting the types of institutions offering adult education and training programmes. Then I present some evidence of the take-up of programmes and courses. Finally, I present some evidence on the results of the inputs. Another dimension of skills in a wide sense concerns health. I give a brief review of recent reforms in the Swedish system of rehabilitation and also present some evidence on the effects. 3.1 Upper secondary vocational programmes Before turning to adult education and training, it is useful to give a brief review of vocational programmes in Swedish the upper secondary school. Vocational programmes in Sweden have a number of characteristics: First, programme dimension is in principle governed by student demand. One might conjecture that this plays a role for a second feature of the system a small content of training in workplaces (and only a tiny fraction of each cohort in apprenticeship programmes). There is no guarantee that student demand for programmes (which is based on, at best, fragmented knowledge of the labour market outcomes associated with different programmes) is matched to employer interest in offering slots for in-workplace training. Third, a reform in the 1990s involving an increased academic content in the vocational programmes primarily led to increased dropping out (Hall, 2012). Finally, stricter eligibility requirements have gone hand in hand with a large and increasing fraction of each cohort leaving compulsory school not 12 IFAU - Employment outcomes and policies in Sweden during recent decades

13 eligible for any national programme. Most of the ineligibles would previously have entered national vocational programmes. Summing up: Sweden has an upper-secondary vocational programme with a small content of in-workplace training, a fairly large academic content and a significant fraction of young persons who end up outside the system because they fail to meet eligibility criteria. 3.2 Adult education (Komvux and Sfi) The Komvux adult education system is designed to offer education to adults at least 20 years old without finished primary or secondary school. Courses are also open for persons below age 20 who have finished upper secondary school, but only in subjects where they have not reached a pass grade. The system is run by the municipalities and offers courses free of charge for the participants. 7 Participation also provides eligibility to public study grants and loans. Immigrants above age 16 can attend courses in Swedish for immigrants (Sfi), which are supposed to provide basic knowledge of the Swedish language and society. The courses are free of charge and can be combined with other activities such as work, practice or education. 3.3 Adult vocational training programmes (Yrkesvux) Originally, Komvux did not include vocational training. However, at the time of the international financial crisis, in 2009, the government decided to give the municipalities the opportunity to apply for funding for vocational courses, originally called Yrkesvux. Priority should be given to students with low educational attainment or to unemployed. The courses offered are often similar to those offered in vocational labour market training provided by the Public Employment Service (PES), see Section 3.5. Municipalities and the PES often buy the courses from the same providers. 3.4 Vocational introductory jobs (Yrkesintroduktionsanställningar) A few years ago (early in 2014), a programme called Vocational introductory jobs (Yrkesintroduktionsanställningar) was launched. The programme combines work and vocational training (the training part at most 25 % of a full time). The programme is 7 The municipalities receive central government money to organize the courses, which are typically bought from private providers. IFAU Employment outcomes and policies in Sweden during recent decades 13

14 targeted, first, at youth lacking experience of a vocation or unemployed at least 90 days, second, at persons at least 25 years old and long-term unemployed or, third, at recently arrived immigrants. The government pays a wage subsidy (corresponding to the payroll tax) and also pays for the vocational training. The programme is based on collective agreements at the central (sectoral) level, specifying the type and amount of training, and this set-up is the result of three-part negotiations between the government and the social partners. The programme can be viewed as an embryo to apprenticeship programmes, but has so far reached (very) modest volumes; far below the guesstimates in official policy documents. 3.5 Vocational labour market training programmes (Arbetsmarknadsutbildning) Vocational training programmes for the unemployed arranged by the Public Employment Service (PES) have a long history in Swedish active labour market policy. Unemployed job seekers are assigned to training programmes by case workers at the PES, but the courses have been procured from private providers since the 1980s. The courses are typically not longer than six months, but typically longer than a few weeks (the Danish system). The general idea is to identify skills that can be acquired in the relevant time span and that are in demand (bottlenecks; occupations in shortages). The participants receive economic support corresponding to the level of unemployment benefits. 3.6 Vocational college (post-secondary) programmes (Kvalificerad Yrkesutbildning and Yrkeshögskolan) Vocational college programmes (Kvalificerad Yrkesutbildning; KY) started as a pilot in 1996 and became a regular programme in In 2009 KY and a few other vocational programmes became Yrkeshögskolan (YH). The objective of the programme has, since the start, been to contribute to a general raise in competence in the labour force, but it has more specifically been an objective to prepare the labour force for structural change due to technical change. The YH programme is broadly targeted: individuals in the labour force needing further training, individuals who want to change occupation and new graduates from secondary education. A distinctive feature of the programmes is that the content is heavily influenced by the social partners: Employers participate in control groups and supply slots for practice and the social partners are involved in the selection of courses and content. 14 IFAU - Employment outcomes and policies in Sweden during recent decades

15 Students seem to find work easily (compared to matched traditional college students), but no full-fledged evaluation is available 3.7 Participation in adult education and training in Sweden Arbetsmarknadsekonomiska Rådet (2018) surveyed Swedish adult education and training from a number of angles. First, they cite survey evidence showing that adult education and training is more extensive in terms of participation in Sweden than in any other country covered by surveys undertaken by Eurostat. This survey-based measure includes both formal and informal education and training. The evidence shows that participation is high in all age groups covered by the study. Participation is higher in Sweden than in other countries both for natives and individuals born abroad, although participation among natives is around ten percentage points higher than among individuals born outside Sweden. The high participation rates in Sweden partly result from high participation in formal programmes, but Sweden really stands out when it comes to participation in informal training. An interesting pattern is that natives have much higher participation in informal training, especially in workplace training, whereas natives have a significantly lower participation rate than individuals born outside Sweden in formal training. The numbers of participants in adult education (including vocational programmes) and Swedish for immigrants have been growing over the past ten years and are substantial in 2017 there were participants in adult education and in Sfi. These figures are substantially higher than the number of participants in vocational labour market training programmes around Participation in labour market training programmes has also been substantially lower in recent years than previously. Also, the vocational college programmes have more participants than labour market training programmes; in 2016 the number of students was around The number of persons with vocational introductory jobs has never been large in May 2018 there were 725 people on such contracts. 3.8 Employment effects of the different programmes In this section I review the evidence of effects of the different types of Swedish adult training and education. Most of the available evidence pertains to labour market training, and there is no evidence on adult vocational training programmes and vocational IFAU Employment outcomes and policies in Sweden during recent decades 15

16 introductory jobs, almost no evidence on vocational college programmes and some but not very much evidence on adult education, including Swedish for immigrants Adult education Some studies have found that adult education has had positive impacts on future earnings, but also that the effects have been of moderate sizes and appear only with a sufficiently long follow-up horizon. 9 It also seems that effects have been heterogeneous: persons with below-average prior incomes have experienced better than average impacts (Stenberg 2010, 2011). Among older participants the estimated effects have been positive for women only (Stenberg et al. 2012, 2014). A few studies have examined the knowledge lift policy initiative in different ways. The knowledge lift was a central government policy initiative during which adult education was expanded substantially and more targeted at unemployed workers. Courses included covered both theoretical and vocational topics. Estimates of the effects for participants are somewhat mixed, but most studies suggest positive impacts, at least in the long run (Albrecht et al., 2005; Stenberg & Westerlund, 2008, 2014). Albrecht et al. (2009) studied general equilibrium effects of the policy initiative, and found both positive treatment effects for the participants and negative (displacement) effects for low-skilled non-participants and positive wage effects for medium skilled non-participants. The estimated overall net effect was positive Swedish for immigrants Swedish for immigrants (Sfi) has been a part of Swedish integration policies for a long time. However, research on the effects of Sfi is very limited. Historically, both participation rates (Kennerberg & Sibbmark, 2005) and completion rates (Statskontoret, 2009) have been fairly low. Evaluations by the National Audit Office (Riksrevisionen, 2008) and Kennerberg & Åslund (2010) give ambiguous conclusions, but possibly suggest positive effects of Sfi on labour-market outcomes for refugee immigrants. 8 The evidence is surveyed in Arbetsmarknadsekonomiska Rådet (2018), Forslund & Vikström (2011) and Calmfors et al. (2004). Notice that much of the evidence concerns effects on future earnings, which only partly reflect employment opportunities and effects on employment. 9 See for example Stenberg (2010, 2011) and Arbetsmarknadsekonomiska rådet (2018). Most studies of adult education look at the knowledge lift policy initiative. The most notable exception is Arbetsmarknadsekonomiska rådet (2018), in which adult education in 2004 is studied. 16 IFAU - Employment outcomes and policies in Sweden during recent decades

17 3.8.3 Vocational labour market training programmes Vocational labour market training programmes have a long history as a part of Swedish ALMPs. A substantial number of studies have estimated (un)employment effects of vocational labour market training programmes. All but the most recent studies are surveyed in Calmfors et al. (2004) and Forslund & Vikström (2011). Arbetsmarknadsekonomiska Rådet (2018) also provides a selective survey. There is a striking variation over time in the estimated effects: The estimated effects for participants in the 1980s were large and positive, most estimates for the 1990s were at best nonnegative, 10 estimates from the early 2000s again strongly positive and, 11 finally, estimates for recent years positive, but close to zero. 12 So, even if estimates from time to time suggest strongly positive effects, the estimates are not consistently indicating good effects. Given that the programme is very expensive, this means that it at best probably has survived a cost-benefit analysis only during limited periods of time Some reflections on what evaluations of adult educations programmes have taught us A first important observation in this brief survey of evidence on adult education is that there are many aspects of adult education, the effects of which are largely unknown. But another important problem concerns the external validity of the studies. Most of the studies of the effects of (general) adult education examine a strong expansion targeted at unemployed workers in a deep recession. Is this evidence valid for other groups in other faces of the business cycle? The results in Arbetsmarknadsekonomiska rådet (2018) suggest that they might be. The evidence on the effects of vocational labour market training programmes, however, suggests that effects do vary a lot for this programme. There is no strong reason to believe conditions to be radically different for other programmes. Hence, I think a sceptical attitude to the evidence is indeed warranted. We just don t now very much about the effects, in the labour market or elsewhere, of adult 10 An interesting exception is provided by Vikström & van den Berg (2017), where it is shown that using a longer follow-up horizon and up-to-date methods suggest significantly positive treatment effects also for the 1990s. 11 For example, the estimates in de Luna et al. (2008) suggested that participation decreased the expected time from unemployment to work by almost 25 % for participants A number of explanations to this development over time of estimated effects focus on a number of possible factors: institutional reforms (primarily whether participation in training programmes gave renewed eligibility for UI benefits or not), programme volumes and changes in volumes, auxiliary targets (especially the so-called 70%-target, according to which at least 70% of the participants should have a job within three months after programme completion), business cycle factors. Some of these factors are discussed in Calmfors et al (2004) and Forslund & Vikström (2011). 13 This is suggested by the analysis in Forslund et al. (2013), where the results indicate that although vocational labour market training programmes outperformed practice programmes, they were too expensive to bear the extra costs in that comparison. IFAU Employment outcomes and policies in Sweden during recent decades 17

18 education measures (with the possible exception of vocational labour market training programmes 14 ). 3.9 Measures of skills among adults and adult education An alternative route to gain knowledge of the effects of adult education is to look for direct measures of skills among the adult population. There is no abundance of such measures, but the OECD has conducted three interesting studies of skills in the adult population in selected member countries, two of which Sweden participated in: the IALS (International Adult Literacy Survey; ) and the PIAAC (Programme for International Assessment of Adult Competencies; ) studies. In the IALS we see that Swedish adults perform really well 15 and that this is most pronounced in a comparison of older persons. Young Swedes also do well, but relatively not as well as the older. We also see that Swedes with less than secondary education do very well. Of course, it is by no means clear that these findings reflect adult education. Also, in the PIAAC study Sweden ranks high, but no longer as number one (possibly with the exception of the measure of problem-solving skills using ICT, where international comparability is more limited). An analysis in Arbetsmarknadsekonomiska rådet (2018) shows that there is a strong correlation between participation in adult education, in particular work-related adult training, and measured skills in PIAAC in a number of countries. Sweden is no exception. However, an attempt to check whether this also reflects a causal relation fails to produce significant results. 16 Hence, neither the use of direct measures of skills gives rise to conclusive evidence on the effects of adult education and training Rehabilitation The rehabilitation chain The Swedish sick-leave process was fairly unstructured for a long time. A reform package issued in 2008 was intended to change that. The reform launched as a so-called rehabilitation chain, the main content of which was that those on sick leave for extended periods should undergo a process with actions taking place no later than at a number of 14 Notice, however, that the recent drop in estimated effects of the vocational labour market retraining programme has been notoriously difficult to explain, see Liljeberg (2016). 15 Sweden ranks as number one in all measured dimensions. 16 Although point estimates are large. 18 IFAU - Employment outcomes and policies in Sweden during recent decades

19 fixed dates. During the first 90 days of a sickness spell, a person s work capacity should be tested against his/her ordinary job. Between days 90 and 180, work capacity should be tested against any job at the employer. Between days 180 and 365, work capacity should be tested against any normal job in the regular labour market unless a number of conditions would say otherwise (one such condition was that the person could be expected to be able to return to the old job before day 366). After day 365 work capacity should be tested against the whole labour market unconditionally. Hägglund (2012) and Hägglund & Österlund (2015) studied the rehabilitation chain and found that the time limits contributed to a more rapid return to work, but no longterm effects on labour supply Work-life oriented rehabilitation During the whole rehabilitation chain, the Social Insurance Agency (SIA) has an overall responsibility for work-life oriented rehabilitation, but the process also involves employers, occupational health care and the PES 17. In individual cases, the responsibility of the SIA is to investigate the conditions for rehabilitation and decide on the eligibility for sickness benefits. This is done through contacts with the parts involved and possibly meetings, the aim of which is to coordinate measures and rehabilitation for a rapid return to work. The employer is responsible to ensure that what is needed for an efficient rehabilitation is done, but only for such rehabilitation that aims at return to work. Sanction may be imposed on employers failing to meet the obligations, but such sanctions are seldom imposed (Johansson et al., 2010). In addition, the employer pays only 14 days of sick-pay. Hence, there are no strong economic incentives for the employer to be active in the rehabilitation process. The rehabilitation chain was complemented by a rehabilitation guarantee in The guarantee was the result of an agreement between the government and Sweden's municipalities and county councils (SKL). Within the guarantee two treatments, cognitive behavioral therapy (CBT) for individuals with mild or moderate mental illness and multi disciplinary treatment (MDT) for individuals with pain in back and shoulders, were offered. The treatments should be offered following an examination and diagnosis undertaken in the health-care system. 17 Work-life oriented rehabilitation is discussed in detail in Johansson et al. (2010). IFAU Employment outcomes and policies in Sweden during recent decades 19

20 An evaluation (Hägglund et al., 2014) found that CBT decreased sickness absence for persons who were not on sick leave when treatment started but had no effect for individuals on sick leave when the treatment stared, while MDT increased sickness absence. Engström et al. (2017) used an RCT to study the impact of early interventions in sickness absence spells. The aim of the interventions was to screen and, if necessary, to rehabilitate sickness absentees. The main finding was that the early interventions, perhaps contrary to expectations, increased the inflow into disability benefits by around 20 percent. The main conclusion in a survey of earlier studies of the effects of rehabilitation in Johansson et al. (2010) was that there is very limited support for work-life oriented rehabilitation as a way to reduce sickness absence. Hence, there is little to suggest that initiatives in sickness insurance and rehabilitation have played a major role for the development of the Swedish employment rate Policies for immigrant integration In December 2010, the responsibility of integration policies was moved from the municipalities to the PES. Since then, the pivot of policies for immigrant integration is a two-year introduction programme at the PES. This relocation of responsibility was evaluated by Andersson Joona et al. (2016), finding a positive impact on employment and earnings. One should, however, be careful when drawing conclusions about the labour market effects of this rather comprehensive reform. As part of the introduction programme, introduction guides were introduced in December In the programme optional activities could be offered newly arrived immigrants in the introduction programme. The guide was supposed to give the immigrants support to facilitate integration. The system does not seem to have worked well (Riksrevisionen, 2014; Sibbmark et al., 2016) and was abandoned. The large recent inflow of asylum seekers has also triggered a number of new policies. One such policy is so called fast tracks to employment for newly arrived immigrants. These fast tracks aim at transferring refugees with relevant skills and experiences to occupations where employers face difficulties in finding the right competence. Information on the actual content of the different fast tracks is scarce, and as yet very few refugees have actually entered them. What seems to be clear is that validation should be 20 IFAU - Employment outcomes and policies in Sweden during recent decades

21 an important component, and it is also clearly stated that the procedures should be specific to each track and that the social partners (employers and unions) should play an active role. However, as yet there are no evaluations of the fast tracks available. The same, unfortunately, holds for validation of informal skills Skills upgrading institutions and policies a brief summing up Are skills upgrading institutions and policies likely to be main drivers of the high Swedish employment rates? Probably not. There is no empirical evidence available to suggest that skills upgrading has played any major role for labour market outcomes. Partly this statement reflects the fact that we simply do not know that much, but also that the relevant empirical evidence does not suggest any consistent and major impacts of the measures that have been evaluated. Looking at the stylised facts of Swedish labour market outcomes I presented in Section 2, we saw that Sweden stands out primarily in terms of high labour supply and not in terms of low unemployment rates. Even though skills upgrading may have an impact both on labour supply and unemployment, it is arguably the case that we should expect skills upgrading to impact more on unemployment than on labour supply (at least in a quantitative sense). 18 Hence, in this perspective, the available empirical evidence on the modest effects of skills upgrading policies is consistent with the fact that Sweden does not stand out in terms of low unemployment rates. This is, of course, not to deny that skills upgrading may have other important impacts, e.g., on productivity growth. 4 Job-search incentives Job search (or labour supply) is potentially affected by a large number of institutions. In principle, everything that has an impact on either costs or benefits associated with job search will affect job-search behaviour. Generally, we would expect that factors that make work more and non-work less beneficial to increase job-search efforts (and, hence, employment). We also expect that factors giving rise to increased search efficiency or lower search costs normally would give rise to higher employment. However, these measures will be discussed briefly in a later section on matching. 18 With the possible exception of rehabilitation measures. IFAU Employment outcomes and policies in Sweden during recent decades 21

22 Then, a number of factors are obvious candidates for the role of influencers of employment: income taxes, unemployment benefits, social assistance benefits, pensions, early retirement schemes, child care, parental leave systems, job-search assistance including information systems. In this section I will survey reforms affecting these factors (except job-search related measures) and empirical evidence on effects of a few of the reforms. 4.1 Income taxes It is well known that Swedish taxes have been high. The labour-supply incentives of taxes reflect both marginal and average tax rates. This means that a given amount of taxes collected will affect labour supply differently depending on the design of the tax system. The Swedish income tax system is divided in two parts; a proportional tax imposed by local governments and a slightly nonlinear national (state) income tax. The local tax rates differ, with an average rate (the average of the sum of the county, municipal and church tax rates) around 30 per cent. However, recent tax reforms have in effect equalised local taxes paid (as opposed to local taxes received by municipalities) through state transfers, see below for more details on these reforms. Income taxation has undergone many reforms during the last four decades. A major reform was implemented in 1971, when in effect household taxation was replaced by individual taxation. As income taxes were progressive, this reform implied lower marginal tax rates for married women. 19 In the early 1980s, the national income tax was highly progressive with many tax brackets. Summing local and state taxes, an average earner may have faced a marginal income tax rate of around 50 per cent, and the top rate exceeded 80 per cent. In marginal income tax rates for average and high-income earners were reduced as was the number of tax brackets. In 1991, another major tax reform was implemented, resulting in a much simpler system with two tax brackets; a zero state income tax for taxable incomes below a threshold and 20 per cent for higher incomes. In 1995, however, an additional tax bracket was introduced, meaning that the top rate for the state income tax since then has been 25 per cent. The next substantial reform initiatives took place during the Reinfeldt governments in five steps between 2007 and The reform initiatives can be described as an implementation of an earned 19 Men typically were primary bread-winners in Swedish households in IFAU - Employment outcomes and policies in Sweden during recent decades

23 income tax credit (EITC) equal for all individuals below age 65, regardless of marital status or number of children in the household. Also, the taxpayer does not need to apply for the tax credit. Finally, there was no phase-out region. 20 The main stated objective was to stimulate labour supply, especially at the extensive margin. Comparing marginal and average tax rates before and after the reforms, one can distinguish a number of features of the reforms. First, it is clear that the policy implied substantially lower average tax rates at the bottom of the income distribution. Second, at higher income, especially at incomes above the point where the EITC is fully phased in, the gap between the average tax rate with and without EITC becomes very small. Third, marginal tax rates at low income levels were substantially reduced and rates at most medium income levels were lowered, whereas marginal tax rates at high income levels remained unchanged. Most empirical evidence suggests that income taxes have had rather modest labour supply effects, at least on the intensive margin. 21 From an employment perspective, however, the interesting margin is the extensive one. Selin (2014) estimated large participation effects of the 1971 change from household to individual income taxes: according to the results, the employment rate among married women would have been ten percentage points lower in 1975 under the old income tax system. The recent EITC reforms have been hard to evaluate with convincing identification strategies; the only serious attempt to estimate the causal effects of the EITC on employment concluded that the variation available to identify effects was insufficiently large to produce reliable estimates (Edmark et al., 2016). On an assignment from the Swedish parliament, the Swedish Ministry of Finance performed a careful evaluation of the EITC where many kinds of evidence were used to reach an overall conclusion about the (un)employment effects of the reform initiatives (Proposition 2011/12:100, Bilaga 5). The main conclusion was that the reform package had contributed significantly to lower unemployment and, especially, to higher employment (between and extra jobs in the long run) There was also an EITC targeted at individuals above age 65. I come back to this when I discuss measures targeted at increasing employment among the elderly. 21 Empirical evidence is surveyed in Aronsson & Walker (1997, 2010). It should be mentioned that Ljunge & Ragan (2004) estimated relatively large effects, but on the tax base rather than on labour supply. 22 The assessment that employment was affected more than unemployment by the reforms reflected a judgement that some of the increased labour supply induced by the reforms pertained to groups with below average job-finding rates. IFAU Employment outcomes and policies in Sweden during recent decades 23

24 4.2 Unemployment insurance A generous unemployment insurance (UI) with a low level of monitoring weakens jobsearch incentives. This can be expected to lead to a lower job-finding rate at given wages, higher wages and, hence, a lower employment rate in equilibrium. 23 Hence, reforms affecting the generosity of UI or the level of monitoring can be expected to have an impact on employment. The Swedish UI is somewhat special. The rules for UI are regulated by law and the UI system is predominantly tax financed. At the same time the insurance is run by a number of independent organisations, UI funds, most of which have (or have had) close ties to trade unions. Membership, which is necessary for income related unemployment benefits, is voluntary. Until recently, membership was close to 90 % of the labour force, but membership declined rapidly by approximately ten percentage points after reforms about ten years ago which led to substantially higher membership fees. Eligibility for income-related benefits requires membership in a UI fund for at least 12 months and in addition a work-history requirement stipulating some minimum amount of work prior to the unemployment spell. The unemployed must also be job ready, actively looking for a job and prepared to accept a suitable job offer. Job-search behaviour is monitored by case workers at the public employment service (PES) and violations of the rules may lead to sanctions. Unemployed who are not members of a UI fund, but meet the same work and job search requirements, are entitled to a fixed daily benefit, which is substantially lower than the maximum income-related benefit. Benefit eligibility lasts for 300 benefit days (420 calendar days; benefits are paid five days a week). The maximum replacement rate is 80 % during the first 200 benefit days and then 70 % until day 300 of the unemployment spell. After this period, the unemployed gets 65 % of previous income conditional on entering an activation programme. The system is, however, much less generous for most unemployed, because there is a cap on the daily benefit which is at a fairly low level. On the other hand, a number of supplementary insurance programmes have been established through collective agreements between trade unions and employers organisations. 23 This conclusion would, for example, follow from a standard Pissarides-Mortensen equilibrium search and matching model. 24 IFAU - Employment outcomes and policies in Sweden during recent decades